In the manufacture and assembly of magnetic media storage disks for storage drives (e.g., hard drives), a number of disk defects are commonly found. These defects may require debugging that can be both time consuming and expensive to perform. A failed disk of this sort will often have to be removed from a failed drive and subjected to media certification or testing to determine if the magnetic disk is usable or not. The media certification determines the number of defect counts in a given washer shaped area of the disk. If the number exceeds a predetermined threshold, then the magnetic disk is put aside for further processing. If the number falls below the predetermined threshold, then the magnetic disk is recycled into the hard disk drive manufacturing process.
A media test tool such as the MC900 (media certifier 900 series) can be used to perform the media certification. However, a known short-coming of this tool and other such tools in the industry is the ability to detect spiral/circular scratches and light micro scratches. As such, an alternative method for detecting and screening out such defects is needed.
Pattern recognition is an established science that is sometimes applied to detecting defects on media. Pattern recognition employs complex mathematical modeling to perform feature extraction, feature recognition and then assimilation of the results for final classification. While aspects of pattern recognition are relatively well established, it is believed that such techniques have not been successfully applied to address the detection of a number of defect types including both light micro scratches and spiral/circular scratches.